1. Field of the Invention
The present invention relates to a video coding/decoding apparatus used in a video communication system such as a videophone, a video conference system, a system of a storage apparatus such as a digital VTR, a video disk, a broadcast system such as a terrestrial, a satellite, a CATV system.
2. Description of the Related Art
In each field of the broadcast system, the storage apparatus, and the video communication system, importance of a video coding for compression-coding a video signal with high efficiency has been considerably increased. In the filed of the broadcast system, there have been advanced the study and development in which the CATV broadcast, the satellite broadcast, and the terrestrial broadcast are digitized so as to realize the multi-channel broadcast, and diversity of service. In the field of the system of the storage apparatus, there have been advance the study and development of the digital VTR and digital video disk, which are applied to the present TV and a HDTV in the future. In the field of the video communication system, there have been advanced the study and development for realizing a multi-media communication through a broadband communication network. Moreover, in the field of the video communication system, a multi-media communication using a mobile radio communication network has been also studied. The former communication system is used in a relatively wide band. In contrast, the latter communication system is used in a finite radio band. Due to this, it is required that an amount of data be largely compressed in the transmission of the video signal.
In recent years, in the field of the video coding, there has been a trend in improving quality of prediction of a motion compensation prediction coding, which is currently performed by the parallel shift of blocks, in order to improve a compression ratio, that is, coding efficiency. One of these new techniques of motion compensation is called a patch motion compensation or a motion interpolation. More specifically, in this new the motion compensation, a motion vector corresponding to each pixel is calculated from motion vectors obtained for sample points by a linear interpolation (see "Basic Study on Motion Compensation with Triangles" by Nakaya, Harashima, Technical Report of IEICE, IE90-106, G. Sullivan, R. Baker, "Motion Compensation for Video Compensation Using Control Grid Interpolation" ICASSP '91, M9.1). According to these methods, since the motion can be set by a pixel unit even if the motion vector having the same amount of data as the conventional motion compensation coding system is sent, a predictive precision can be improved, and this can contribute to the compression ratio of video data. Moreover, since the basis of the conventional motion compensation coding system lies in the preparation of a predictive video signal by a block unit, there is a tendency to generate blocking artifacts in a decoded video signal. According to these methods, the predictive video signal is continuously generated, and no inter block gap in the video content is generated.
On the other hand, as a general technique in the video coding, it is needed that a coding output amount of an coding apparatus be smoothed, that is, a rate control for adjusting the coding output amount to a target coding output amount is required. According to the conventional video coding apparatus, a quantization step size is changed as viewing a storage amount (amount of occupation) of an output buffer so as to control the coding output amount. However, if there is used an orthogonal transform coding such as H. 261, which is an international standard system of video coding, and DCT used in MPEG, there is a problem in that the blocking artifacts are generated to largely deteriorate a subjective quality of video.
In order to solve the above problem, there is known a rate control using a low-pass filter, for example, "Review of Variable Pre-filter Control System in Video Coding" by Takishima, Wada, PCSJ'92 6.6). This technique explains that not only the quantization step size but also resolution of an input video signal before being coded is reduced so as to control the coding output amount. As compared with the rate control in which the coding output amount is controlled by only the quantization step size, resolution of the large motion video drops, but the blocking artifacts can be reduced, so that the subjective quality can be improved.
Among the above-mentioned prior art, regarding the patch motion compensation and the motion interpolation, the prediction quality in the small motion of the video can be improved, and the coding efficiency can be also improved. However, regarding the large motion of the video in which motion compensation cannot be performed, influence of its deformation is exerted on the large area. As a result, there may be a case that a prediction error is increased, and the coding efficiency is reduced. Particularly, in the case of a change of a scene in which the motion compensation does not originally work, no improvement is made.
The motion vector is normally detected by comparing the input video signal and the decoded video signal. Due to this, in a case where the coding is performed at a low bit rate, a suitable motion vector cannot be easily detected by deterioration of the decoded video signal due to influence of coding noise, unfavorable influence is exerted on the predictive video. Moreover, for obtaining the motion vector suitable for the patch motion compensation and the motion interpolation, the motion vector, which is obtained by the block matching, which is the conventional vector detection system, is set as an initial value, and the motion vector is optimized by an iterative operation. Due to this, the amount of calculations are largely increased.
On the other hand, in the rate control using the low-pass filter, since the filter is passed over the entire screen, there is a problem in that deterioration is prominent when resolution largely drops. Particularly, in a case where the large motion and deformation are formed in only a part of the area of the same video, resolution of the small motion area where the coding output amount does not become large is dropped. Due to this, the subjective quality is caused to be reduced as visual deterioration.